# temperature sensor

I wish to measure the temperature of the an enclosed box, the range would be from 0C to 80C. And I want the measurement to be pretty accurate. If the real temperature is 25C I want me reading to be +/- 1C.

So I went to RS and did a quick search.

• Maxim DS18B20+ ±0.5°C AUD 5.87

My question is why AD590 cost a lot more with less accuracy? For my requirement will Maxim do the job? Is that any catch? Or any suggestion, preferable through hole.

• If the real temperature is 25C I want me reading to be +/- 1C ... what accuracy do you want at other temperatures? – jsotola Apr 17 '18 at 7:26
• @jsotola +/-1C through out the range that I want. – Sam Apr 17 '18 at 7:28
• It just maybe 'chance' but for a good comparison there are other parameters like factory calibration, long term stability, dependence on power supply noise, digital interface interference. I do know that in the good old days if I needed a top-quality analogue chip I would first go to AD. – Oldfart Apr 17 '18 at 7:32
• do you have anything against using a thermistor? – IC_Eng Apr 17 '18 at 8:46
• @IC_Eng no, but I don't have references to calibration NTC – Sam Apr 17 '18 at 9:46

IC temperature sensors have a wide price range. Older chips tend to keep the price, the manufacturers don't feel the need to lower it.

Note that some are much better than the specs especially a 1% spec (but some are worse)

Precision NTC thermistors are very good.

Here's the first one on Digikey: 0.05C 0-50C for \$7

They have good interchangeability of (holy shit 0.05C - used to be 0.3C) degrees, very low noise, very easy to measure accurately (voltage reference not needed), and low thermal leakage through the fine wires.

But, you have to calculate the temperature.

I am also fond of the IC's that use transistors as diodes with resistance cancellation like TMP513,ADM1034. You just use any bipolar transistor as the sensor, and they are remarkable interchangeable. The nice thing is that you can get transistors in any package/size you can imagine, and they cost next to nothing

Note that when you use digital sensors with internal sensing, the conversion power results in systematic error.

Also be aware that apart from accuracy, some actually have very high noise levels - significantly greater than the resolution they boast.

The AD590 is a very old established chip. Pricing does not necessarily have to be logical, and IC temperature sensors are rather limited in application. This particular part has been designed into a lot of expensive applications and folks are loath to change. For example, the AD590 is used for compensating navigational systems in space and missile applications and it has adequate radiation hardness etc. for that kind of application.

Thermistors and RTDs are often better choices for temperature ranges around room temperature. Thermistors have high sensitivity, are very nonlinear, and high initial accuracy is available. Stability is potentially less, particularly at higher temperatures. RTDs are very stable (in fact they are the lab standard way of interpolating between fixed points on the temperature scale) but have less sensitivity. Obviously they don't put a part like an AD590 into consumer goods. The 18B20 has a digital output so your error budget calculations are less complex (with an analog sensor you have to consider the ADC, reference for the ADC etc). In either case, you should use a an official supply chain for procurement. There are many other such sensor chips with various advantages and disadvantages.

It is not clear whether you are trying to measure air temperature or PCB temperature, or something else, and what response time you require. A sensor from any reputable company should relay the die temperature of the sensor to you in accordance with the datasheet specifications, but the die may not follow whatever you are trying to measure very well. All sensors have some internal heating, for example.